US AI Executive Order (October 2023)

Overview

Executive Order 14110, signed by President Biden on October 30, 2023, was the most comprehensive US federal action on AI. It established new standards for AI safety, directed federal agencies to take specific actions, and addressed competition, privacy, civil rights, and workforce impacts.

Note: The status of this EO under subsequent administrations should be verified — executive orders can be revoked or modified by new presidents.

Key Provisions

Safety & Security

• AI systems posing risks to national security, economic security, or public health must report safety test results to the US government
• Dual-use threshold: Models trained with >10^26 FLOPs (or equivalent biological risk capability) must report to the government before release
• NIST to develop standards and red-team testing frameworks
• Watermarking and content authentication standards for AI-generated content

Privacy

• Federal agencies directed to evaluate privacy-preserving AI techniques
• Calls for federal privacy legislation (Congress hasn’t acted)
• Assessment of how AI increases risks of surveillance

Equity & Civil Rights

• Guidance to prevent algorithmic discrimination in housing, lending, employment
• Federal agencies to address AI bias in their own use of AI

Competition & Innovation

• Support for small businesses and startups
• Immigration reforms to attract AI talent
• National AI Research Resource (NAIRR) for academic researchers

Workers

• Study AI’s impact on the labour market
• Develop best practices for AI in workplaces
• Support workers displaced by AI

Comparison: US vs EU Approach

Aspect	US (EO 14110)	EU (AI Act)
Legal force	Executive order (not law)	Regulation (binding law)
Approach	Voluntary + reporting	Mandatory compliance
Risk framework	Dual-use focus	Risk-based categories
Enforcement	Agency guidelines	Fines up to 7% revenue
Scope	Federal only	All EU member states
Flexibility	Can be revoked	Requires legislative repeal

Key Implications

• Not a law — can be modified or revoked by future presidents
• Reporting requirements created a de facto registry of frontier AI systems
• Influenced industry voluntary commitments
• Highlighted US preference for “light touch” vs EU’s regulatory approach

Questions / Follow-up

• What is the status of this EO under the current administration?
• What state-level AI laws have been passed? (California, Colorado, etc.)
• How do the NIST AI Risk Management Framework and ISO/IEC 42001 relate?

Resources

• Full text of EO 14110
• NIST AI Risk Management Framework
• AI.gov — US government AI resource hub